Coordinates three specialized AI agents to review, implement, and vote on code changes.
npx skills add https://github.com/Enderfga/three-minds --skill three-minds
Please help me install this Agent Skill. Skill: three-minds Source: https://github.com/Enderfga/three-minds/tree/main Command: npx skills add https://github.com/Enderfga/three-minds --skill three-minds
It solves the problem of getting reliable outcomes for complex code and design tasks by coordinating multiple specialized AI agents that can read and modify the same codebase and iteratively converge on agreement.
Run three-minds with your task and a target project directory, for example: three-minds "Review all code in src/ and propose fixes, then implement improvements" --dir ./your-project.
High-intent search queries and answers about installing and using this skill.
Multi-agent consensus improves code refactoring by coordinating three specialized AI agents that sequentially review, implement, and vote on changes using a shared working directory. This iterative process ensures reliable outcomes for complex design tasks by requiring a final agreement.
To run an automated code review, execute the command with your task description and target project directory, for example: three-minds "Review all code in src/ and propose fixes, then implement improvements" --dir ./your-project.
Yes, you can use different AI models for repo collaboration. The framework routes Claude, GPT, and Gemini through a consistent agent backend with optional proxy support, allowing model switching via configuration and a custom proxy baseUrl.
Yes, multi-agent code review supports automated test execution. The collaboration loop applies to scenarios requiring iterative proposals, file-based edits, and test execution across multiple rounds to validate the implemented changes before reaching a final consensus.
The best way to coordinate specialized agents on the same codebase is using a preset multi-agent collaboration workflow. Agents sequentially handle review, implementation, and voting, enforcing end-of-response voting with a specific consensus tag to confirm agreement.
You should not use a consensus-driven approach for simple, single-step code modifications. The multi-agent loop is designed for complex code review, refactoring, and feature design tasks that benefit from iterative proposals and multi-round consensus before applying changes.
Tx1207Create a new project from a template, initialize uv, and configure a local Git repository.
berrzebbSquash-merge a git worktree into the target branch with a structured commit message.
greencmCompares AI models on Claude Code skills by latency, tokens, cost, and quality.
THUFanZdAutomate multi-round research reviews with a secondary Codex agent.
